Intel 18A Shows Strong Progress, May Power Switch 3 GPU

[Daniel Nenni]

Intel 18A process node is showing significant technical progress and may soon play a critical role in powering next-generation graphics processors, including a possible design win for Nintendo's upcoming console, according to a new report by KeyBanc Capital Markets analyst John Vinh.

Vinh notes that Intel 18A is expected to debut in the second half of 2025 with the launch of Panther Lake, a CPU series leveraging this advanced node. Key process indicators (KPIs) such as yield and defect density are reportedly trending in a favorable direction and have reached acceptable levels.

In late February, Intel officially introduced the Intel 18A process on its website, highlighting its readiness for production. The technology integrates RibbonFET gate-all-around transistors for precise current control and introduces PowerVia, the industry's first backside power delivery system. This innovation boosts transistor density and cell utilization by 5–10%, reduces power delivery resistance, and improves ISO power performance by up to 4% compared to traditional front-side power architectures.

Compared to Intel 3, Intel 18A delivers 15% higher performance per watt and 30% higher transistor density. Intel describes it as the first sub-2nm node manufactured in North America, offering a geographically diversified production alternative for chipmakers.

Independent benchmarking by TechInsights places Intel 18A ahead of its competitors in performance at the 2nm level, scoring 2.53, compared to TSMC N2 at 2.27 and Samsung SF2 at 2.19. This suggests Intel 18A currently leads in the race for next-gen process technologies.

In March, Intel engineering lead Pankaj Marria celebrated a milestone on LinkedIn, announcing that the first production batches using the Intel 18A node were successfully completed in Arizona—marking a major step forward for Intel's advanced manufacturing roadmap.

Based on Vinh's latest findings, the 18A process has achieved stable yields, setting the stage for the timely launch of Core Ultra 300 “Panther Lake” CPUs later this year.

Notably, Reuters previously reported that NVIDIA and Broadcom are actively conducting test runs on Intel 18A, while AMD is also evaluating the process for future designs. Vinh's report goes further, suggesting Intel may have secured a GPU order for Nintendo's future Switch 3 console, leveraging its 18A node for manufacturing.

It's worth mentioning that Nintendo's newly announced Switch 2, unveiled on April 2, is powered by a custom NVIDIA processor featuring RT and Tensor cores for enhanced visual effects and AI capabilities. NVIDIA claims the new system delivers up to 10x graphics performance over its predecessor, with shipments expected to reach 15 million units in 2025.

If Vinh's prediction proves accurate and Switch 3's GPU moves to Intel 18A, ongoing testing by NVIDIA makes perfect sense.

Vinh also observed that Intel has reduced pricing for its Lunar Lake CPUs by 20% to 40% to gain market share, though this aggressive pricing is impacting its gross margin recovery, currently reported at 34.32%. Despite these pressures, Vinh maintains a positive outlook for Intel, driven by the strong progress of Intel 18A and resilient server demand, particularly in the cloud computing and AI segments.

Meanwhile, Vinh downgraded his rating on AMD, citing sustainability concerns in its AI GPU business in China and rising competitive pressure from Intel's aggressive pricing. He believes AMD has limited opportunity to gain market share at Intel's expense given the latter's process advancements.

Interestingly, market demand for AMD's upcoming MI308 GPU remains strong. According to J.P. Morgan analyst Harlan Sur, AMD's AI GPU revenue could grow 60% in 2025.

Intel 18A Shows Strong Progress, May Power Switch 3 GPU - Electronic Components Distributor - SMBOM.COM

Intel 18A process shows strong progress and may secure Nintendo Switch 3 GPU order, marking a major leap in 2nm chip performance and advanced manufacturing. Electronic Components Distributor

www.smbom.com

 

[XtorTourist]

Given the 8 year gap between the Switch 1 and Switch 2 (launch of June 2025, Lord Tariff allowing), I don't think that saying that Intel "may secure" the Switch 3 GPU even meets the level of "speculation."

 

[fansink]

Intel 18A process node is showing significant technical progress and may soon play a critical role in powering next-generation graphics processors, including a possible design win for Nintendo's upcoming console, according to a new report by KeyBanc Capital Markets analyst John Vinh.

Act as if, until it is

 

[siliconbruh999]

Given the 8 year gap between the Switch 1 and Switch 2 (launch of June 2025, Lord Tariff allowing), I don't think that saying that Intel "may secure" the Switch 3 GPU even meets the level of "speculation."

Agreed Switch 3 would launch around 2030 at minimum it's just a fake rumor imo

 

[Mooredaddy]

Intel 18A process node is showing significant technical progress and may soon play a critical role in powering next-generation graphics processors, including a possible design win for Nintendo's upcoming console, according to a new report by KeyBanc Capital Markets analyst John Vinh.

Vinh notes that Intel 18A is expected to debut in the second half of 2025 with the launch of Panther Lake, a CPU series leveraging this advanced node. Key process indicators (KPIs) such as yield and defect density are reportedly trending in a favorable direction and have reached acceptable levels.

In late February, Intel officially introduced the Intel 18A process on its website, highlighting its readiness for production. The technology integrates RibbonFET gate-all-around transistors for precise current control and introduces PowerVia, the industry's first backside power delivery system. This innovation boosts transistor density and cell utilization by 5–10%, reduces power delivery resistance, and improves ISO power performance by up to 4% compared to traditional front-side power architectures.

Compared to Intel 3, Intel 18A delivers 15% higher performance per watt and 30% higher transistor density. Intel describes it as the first sub-2nm node manufactured in North America, offering a geographically diversified production alternative for chipmakers.

Independent benchmarking by TechInsights places Intel 18A ahead of its competitors in performance at the 2nm level, scoring 2.53, compared to TSMC N2 at 2.27 and Samsung SF2 at 2.19. This suggests Intel 18A currently leads in the race for next-gen process technologies.

In March, Intel engineering lead Pankaj Marria celebrated a milestone on LinkedIn, announcing that the first production batches using the Intel 18A node were successfully completed in Arizona—marking a major step forward for Intel's advanced manufacturing roadmap.

Based on Vinh's latest findings, the 18A process has achieved stable yields, setting the stage for the timely launch of Core Ultra 300 “Panther Lake” CPUs later this year.

Notably, Reuters previously reported that NVIDIA and Broadcom are actively conducting test runs on Intel 18A, while AMD is also evaluating the process for future designs. Vinh's report goes further, suggesting Intel may have secured a GPU order for Nintendo's future Switch 3 console, leveraging its 18A node for manufacturing.

It's worth mentioning that Nintendo's newly announced Switch 2, unveiled on April 2, is powered by a custom NVIDIA processor featuring RT and Tensor cores for enhanced visual effects and AI capabilities. NVIDIA claims the new system delivers up to 10x graphics performance over its predecessor, with shipments expected to reach 15 million units in 2025.

If Vinh's prediction proves accurate and Switch 3's GPU moves to Intel 18A, ongoing testing by NVIDIA makes perfect sense.

Vinh also observed that Intel has reduced pricing for its Lunar Lake CPUs by 20% to 40% to gain market share, though this aggressive pricing is impacting its gross margin recovery, currently reported at 34.32%. Despite these pressures, Vinh maintains a positive outlook for Intel, driven by the strong progress of Intel 18A and resilient server demand, particularly in the cloud computing and AI segments.

Meanwhile, Vinh downgraded his rating on AMD, citing sustainability concerns in its AI GPU business in China and rising competitive pressure from Intel's aggressive pricing. He believes AMD has limited opportunity to gain market share at Intel's expense given the latter's process advancements.

Interestingly, market demand for AMD's upcoming MI308 GPU remains strong. According to J.P. Morgan analyst Harlan Sur, AMD's AI GPU revenue could grow 60% in 2025.

Intel 18A Shows Strong Progress, May Power Switch 3 GPU - Electronic Components Distributor - SMBOM.COM

Intel 18A process shows strong progress and may secure Nintendo Switch 3 GPU order, marking a major leap in 2nm chip performance and advanced manufacturing. Electronic Components Distributor

www.smbom.com

Nintendo already said the switch 2 was using a Nvidia soc

 

[dkr1986]

Agreed Switch 3 would launch around 2030 at minimum it's just a fake rumor imo

Switch 1 started development in 2012 & launched in 2017 with Tegra X1 on TSM 20nm then refreshed in 2019 with Tegra X1+ on TSM 16nm. Switch 2 started development in 2019 and is launching in 2025.
TSM 20nm HVM is 2014.
TSM 16nm HVM is 2015?\2016?. (No idea which version of the node is used)

So, I think there is a chance this rumor about Switch 3 being an 18A win could be real based on development time and node process HVM timeline!

Nintendo Switch - Wikipedia

en.wikipedia.org

Tegra - Wikipedia

en.wikipedia.org

Nintendo Switch 2 - Wikipedia

en.wikipedia.org

 

[Daniel Nenni]

Nintendo already said the switch 2 was using a Nvidia soc

I really do not think Intel 18A will be used for low power battery applications. John Vinh is a very smart guy but this one is off.

 

[OneEng]

I really do not think Intel 18A will be used for low power battery applications. John Vinh is a very smart guy but this one is off.

Interesting.

Is this because you believe that Intel's GAA w/ BSPDN is not well suited for this kind of application? I believe I have heard rumor that 18A is more targeted at high performance computing.

I would give you a penny for your thoughts .

 

[Daniel Nenni]

Interesting.
Is this because you believe that Intel's GAA w/ BSPDN is not well suited for this kind of application? I believe I have heard rumor that 18A is more targeted at high performance computing.
I would give you a penny for your thoughts .

Yes, Intel has a performance advantage, TSMC has a power and density advantage. Scott Jones did a full analysis but it did not include packaging.

 

[Artificer60]

Yes, Intel has a performance advantage, TSMC has a power and density advantage. Scott Jones did a full analysis but it did not include packaging.

I wonder if that is the unique niche that Intel should shoot for. When you absolutely, positively must have the most performant chip.... HPC isn't a particularly large volume market, but it does have very high margins.

 

[dkr1986]

Let's not forget 18A-P node which is mobile optimized per Intel. Don't know what is the status of that node (should be ready before 14A) but that could be the node Nintendo wants to use. I guess we have to wait for official announcements.

 

[Daniel Nenni]

I wonder if that is the unique niche that Intel should shoot for. When you absolutely, positively must have the most performant chip.... HPC isn't a particularly large volume market, but it does have very high margins.

I believe the unique niche Intel Foundry is best fit for is the NOT TSMC market. That is the low hanging fruit per say. From what I remember the delta for PPA between Intel 18A and TSMC N2 is single digit. The big value is a second source for leading edge processes and to make that possible companies will need to use it.

I have been asking around and there does not seem to be a design team at Nintendo so they will probably use Nvidia Tegra again. Nvidia is a big fan of TSMC N4 so that is probably what the Switch 3 will use. This is a very high volume product so price is an issue. Just my opinion of course.

 

[siliconbruh999]

I believe the unique niche Intel Foundry is best fit for is the NOT TSMC market. That is the low hanging fruit per say. From what I remember the delta for PPA between Intel 18A and TSMC N2 is single digit. The big value is a second source for leading edge processes and to make that possible companies will need to use it.

I have been asking around and there does not seem to be a design team at Nintendo so they will probably use Nvidia Tegra again. Nvidia is a big fan of TSMC N4 so that is probably what the Switch 3 will use. This is a very high volume product so price is an issue. Just my opinion of course.

Agreed Switch is a value Chip 18A is too advanced and expensive for it.
18A within single digits of N2 is a very good achievement for Intel after the 10nm disaster now it's up to the design/product guys to show what they are capable of cause they are the one that needs to proof their worth.

 

[nghanayem]

I wonder if that is the unique niche that Intel should shoot for. When you absolutely, positively must have the most performant chip.... HPC isn't a particularly large volume market, but it does have very high margins.

Interesting.

Is this because you believe that Intel's GAA w/ BSPDN is not well suited for this kind of application? I believe I have heard rumor that 18A is more targeted at high performance computing.

I would give you a penny for your thoughts .

"HPC" in the foundry world is everything bigger than a smartphone (like 5-8W). In fact, if you look at how TSMC and UMC used to talk about things (and how UMC still talks about things), even smartphones were lumped into the HPC category back when foundries were mostly doing puny little low power <=2W ASICs, microcontrollers, digital watches, calculator chips, etc. But even by modern TSMC terminology, the Switch sits somewhere between a tablet and laptop for TDP and battery capacity and would probably be called HPC considering purpose built tablet SOCs like Apple M series are in the "HPC" bucket. With that said, I wouldn't exactly call the HPC TAM as TSMC defines it (rather than how chip designers often define it) as particularly small. We are talking tablets, PCs, server chips of all flavors (CPU, GPU, networking, ASICs, etc), gaming consoles, modern automotive infotainment, modern automotive ADAS/self-driving chips, and consumer/professional client GPUs. After all in recent times HPC became TSMC's largest end market. Granted AI is hot hot hot; but if we are being honest mobile is stagnant, and even before AI, TSMC's HPC segment was rapidly gaining on the mobile segment earnings report after earnings report.

Agreed Switch is a value Chip

If they wanted "value" they would be using a more advanced chip making process. But I guess after TSMC and NVIDIA all collect their pounds of flesh, maybe this is cheaper than a dieshrink of their design to N4P...

I believe the unique niche Intel Foundry is best fit for is the NOT TSMC market. That is the low hanging fruit per say. From what I remember the delta for PPA between Intel 18A and TSMC N2 is single digit. The big value is a second source for leading edge processes and to make that possible companies will need to use it.

I have been asking around and there does not seem to be a design team at Nintendo so they will probably use Nvidia Tegra again. Nvidia is a big fan of TSMC N4 so that is probably what the Switch 3 will use. This is a very high volume product so price is an issue. Just my opinion of course.

18A is too advanced and expensive for it.

Let my preface by saying, I kind of doubt this rumor. I doubt a choice has been made on what foundry when they don't even know what they would want to do for a device that won't come out for like another 8 years. In 2–4 years, sure they would know since the chip would be in active design or fully locked in depending on where in that range you are. But now, color me skeptical. With that preface out of the way, the closest Nintendo could get to "considering 18A" is looking at how 8LPP came out 7 years before this new device and 18A/N2/SF3/SF2 all would fit a similar maturity level in 8 or so years. But I wouldn't really call that "considering" so much as I would call it, "these are the lagging nodes that would be available when we were looking to buy". Compared to the 1st gen GAA stuff, N4 family would be 13 years old (almost 2x as old SF8 is today). N3E family, I could at least see since it is so close in PPA to 1st gen GAA if the price is sufficiently lower than not just N2 but also 18A and SF3/2 families to justify it. So yeah TLDR is 18A is advanced stuff now, but in 8 years 18A is as ancient and unremarkable as Samsung/TSMC's 10nm process nodes and Intel cannon lake are today.

18A within single digits of N2 is a very good achievement for Intel after the 10nm disaster now it's up to the design/product guys to show what they are capable of cause they are the one that needs to proof their worth.

But silliconbruuuuuuuuuuuuuuh using the manufacturing side as a crutch for your outdated chip designs (even back when you did memory chips), blaming them for every issue you ever had, leaving for TSMC, still making chips that come out late and don't light the market on fire even when using a better TSMC process, and then blaming your old manufacturing partner for the current mess is so much easier and more fun!